Performance evaluation of multi-camera visual tracking

Size: px
Start display at page:

Download "Performance evaluation of multi-camera visual tracking"

Transcription

1 Performance evaluation of multi-camera visual tracking Lucio Marcenaro, Pietro Morerio, Mauricio Soto, Andrea Zunino, Carlo S. Regazzoni DITEN, University of Genova Via Opera Pia 11A Genoa - Italy s: {pmorerio}@ginevra.dibe.unige.it {lucio.marcenaro,carlo.regazzoni}@unige.it Abstract Main drawbacks in single-camera multi-target visual tracking can be partially removed by increasing the amount of information gathered on the scene, i.e. by adding cameras. By adopting such a multi-camera approach, multiple sensors cooperate for overall scene understanding. However, new issues arise such as data association and data fusion. This work addresses the issue of evaluating the performance of a multicamera tracking algorithm based on Rao-Blackwellized Monte Carlo data association (RBMCDA) on real data. For this purpose, a new metric based on three performance indexes is developed. I. INTRODUCTION Video-based tracking techniques are widely used in different fields related to ambient intelligence and scene understanding. Video surveillance is only one of the possible applications ranging from security to man-machine interaction; elderly people monitoring for sanitary assistance or rehabilitation is only one of the interesting fields where video-sensors and automatic processing algorithms can be used efficiently. Moreover, technological improvements in imaging hardware and related price cuts, enabled na even more pervasive usage of such sensors: IP based video cameras allowed easy hardware installation and remote monitoring and configuration while mega-pixel sensors and efficient video compression algorithms greatly improved quality and spatio-temporal resolution of acquired images. In this scenario, visual tracking is one of the principal techniques that are used for extracting high-level descriptors of the scene. The main purpose of tracking algorithms is to estimate moving objects trajectories by correctly assigning them a specific label that must be propagated over-time as the object moves on the scene. Each tracking technique starts from low-level data acquired directly from images, thus can be modelled as a state estimate problem starting from noisy observations. In this sense, tracking can be seen as a filtering process for cancelling observations noise and extracting the actual movement of each object in the scene. Several highlevel applications can be developed based of trajectories and detected movements, such as people counting, traffic monitoring, abandoned objects detection, automotive safety, etc.. The complexity of visual-tracking is very high and main problems that must be solved are related with unpredictable object movements, targets or scene appearance variations, nonrigid parts, occlusions (intra-, inter- or environmental) and sensor undesired movements. A multi-camera approach can be adopted in order to deal with these issues, although data fusion and association problems must be considered when multiple sensors have to cooperate for overall scene understanding. The work is organized as follows: Section II reviews the probabilistic approach to visual tracking and introduces the reader to particle filtering, Rao-blackwellized particle filtering and Rao-blackwellized data association (RBMCDA). Section III describes the experimental set-up and model used for performance evaluation. Section IV introduces the novel metric and performance indexes. In section V performance evaluation results are given. Conclusion are drawn in section VI. Future developments of the work are also proposed. II. PROBABILISTIC APPROACH Visual tracking problems can be mathematically modelled through probabilistic reasoning. Target positions can be stored in a state vector containing its descriptive features: target state must be estimated starting from noisy observations obtained from acquired images. Stochastic state and observation models are needed for this kind of approach. Bayesian theory can be used to estimate the state s a posteriori probability density function (pdf ): after the pdf is estimated, it can be used for evaluating the new state and for measuring the state estimate precision. The hidden state vector to be estimated can be expressed as x t R nx where n x is the state vector dimension and t N is associated to the considered temporal instant. The state vector x t can contain object position, speed, acceleration or shape related features (e.g., corners). Available observation vector z 1:t = {z 1 z t } at time instant t will be used for a posteriori state estimate. An on-line solution is needed for object tracking as one must be able to estimate the new state of each tracked object as new observations become available. The a posteriori state probability density function p(x t z 1:t ) is computed recursively through a two-steps prediction-updating procedure. If the previous step posterior pdf p(x t 1 z 1:t 1 ) is known, the pdf of the following state can be predicted by means of the state transition model p(x t x t 1 ), given by: p(x t z 1:t 1 ) = p(x t x t 1 )p(x t 1 z 1:t 1 )dx t 1 (1)

2 Moreover, by applying the Bayes theorem, it is possible to obtain the a posteriori state pdf p(x t z t ) through the following update equation that shows its dependency by observation probability p(z t z t 1 ) and the likelihood probability coming from the observation model p(z t x t ) p(x t z 1:t ) = p(z t x t )p(x t z 1:t 1 ) p(z t z 1:t 1 ) This formulation assumes that x t is a state of a Markov process with initial distribution p(x 0 ). Equations (1) and (2) have an analytical solution only in the case of linear-gaussian models. The Kalman filter provides the optimal solution. In the case of non-linear or non-gaussian models, a numerical approximation, like the one provided by applying a Particle Filter (PF), must be made [1]. A. Particle filtering The Particle Filter approximates the solution of the Bayesian filtering problem by representing the a posteriori pdf as a finite set of samples (particles) χ t = {x m t, wt m } Ns m=1 that are weighted by using the importance sampling technique. According to this algorithm, a set of N s samples { x m t } Ns m=1 are created starting from a reference distribution q(x 0:t 1 z 1:t ) instead of the a posteriori pdf p(x 0:t 1 z 1:t ) that, in general, cannot be expressed mathematically. The difference between the reference pdf and the a posteriori one is managed through a correction procedure, by using different weights. Under independence hypotheses, the reference distribution can be expressed as q(x t x t 1, z t ) and the weights can be computed by using the Sequential Importance Sampling (SIS) technique as w (m) t w (m) p(z t x t )p(x t x t 1 ) t 1 (3) q(x t x t 1, z t ) According to the SIS technique, N s samples x m t } can be generated starting from the importance distribution q(x t x t 1, z t ) and correspondent weights w (m) t can be computed by using Eq. (3). After a normalization step, the a posteriori pdf at time t can be approximated with a sum of N s weighted Dirac deltas p(x t z 1:t ) δ(x t x (m) t ). (4) w (m) t m=1:n s It has been demonstrated that this algorithm leads to a unstable behaviour due to the continuous increase of the weighting coefficients variance. This can be solved by using re-sampling procedures that try to redistribute particles in such a way that softly weighted ones are cancelled while those associated to higher weights are duplicated. The efficiency of particle filter strongly depends on the appropriateness of the importance distribution q(x t x t 1, z t ) that must be close to the a posteriori pdf. Within visual tracking, the definition of a reliable importance distribution is a complex task because of the high variability of movements and deformations that can be associated to a monitored target. Typically, several features are extracted and tracked from a single target: this problem can be addressed as a multiple (2) target tracking problem for each object. This is a challenging problem because of the uncertainty of the measurement origin (i.e., the centre of the object) and the cluttered and missed detections due to inaccuracies of the feature extractor and/or possible (self)occlusions. It can be demonstrated that data association uncertainty in tracking leads to a NP-hard problem, i.e. the association hypotheses grow exponentially with time as in the Multiple Hypotheses Tracker (MHT)[2]. As a result, approximations are required in order to make the problem feasible for a reasonable number of targets. In the past few years, great focus has been given to sequential Monte Carlo (SMC) strategies under the Bayesian framework because they apply a semi-soft-decision data association approach, i.e. hard-decision associations per particle while keeping multiple parallel hypotheses associations with all particles. Different strategies avoiding explicit enumeration of all possible hypotheses, thus making this approach to scale well with an increasing number of targets, are described in [3]. B. Rao-blackwellized particle filter The algorithm that will be considered hereinafter is an extended version of the Rao-Blackwellized Monte Carlo data association (RBMCDA) [4], which is basically an extended Rao- Blackwellized particle filtering based multiple target tracking algorithm [5]. The algorithm models as stochastic processes not only the state evolution of the target, but also its creation and suppression, i.e. the starting and ending time instant estimations for target tracking are not based on heuristics, but are accounted for in the probabilistic model. The basic idea behind the Rao-Blackwell algorithm is that, due to the conditioning of the data association and the targets creation or suppression, the a posteriori probability distribution of target states can be approximated by means of Gaussian pdf s (marginalization of the filtering or Rao-Blackwellization). Suppose the state x can be split into two substates r and s such that p(x t x t 1 ) = p(s t r t 1, s t 1 )p(r t r t 1 ), (5) and p(s 0:t z 1:t, r 0:t ) is analytically tractable. Decompose the posterior as p(r 0:t, s 0:t z 1:t ) = p(s 0:t z 1:t, r 0:t )p(r 0:t z 1:t ). (6) The term p(s 0:t z 1:t, r 0:t ) can be analytically estimated, while p(r 0:t z 1:t ) can be derived by means of the particle filter as p(r 0:t z 1:t ) = p(z t z 1:t 1, r 0:t )p(r t z t 1 )p(r 0:t 1 z 1:t 1 ) p(z t z 1:t 1 ) (7) This way, target states can be computed, while SMC particle filtering only need to be applied to the data associations and creation and deletion processes. This greatly reduces computational complexity, thus increasing the efficiency of the particle filter. Moreover, the Rao-Blackwell theorem states that the mean squared error of the RaoBlackwell estimator

3 does not exceed that of the original estimator (i.e. of the nonmarginalized particle filter). In this framework, a data association event can be modelled through a data association indicator c t, which takes T t + 1 values: c t = 0 for clutter and c t = j, j = 1,..., T t where T t is the total number of targets at time instant t A priori probabilities for data association can be modelled as a m-order Markov random chain with uniform priors over targets and clutter: p(c t c t 1,..., c t m, T t 1,..., T t m ) = T t 1. (8) Clutter measurement probabilities can be uniformly distributed in a measurement space of volume V, namely p(z t c t = 0) = 1/V. (9) Target measurements and target dynamics are basically modelled as in a Kalman Filter (linear dynamic model and Gaussian noise): p(z t x t,j, c t = j) = N(z t H t,j x t,j R t,j ), (10) p(x t,j x t 1,j ) = N(x t,j A t 1,j x t 1,j Q t 1,j ), (11) Fig. 1. Four different views of the scene where the measurement, covariance, transition and process noise covariance matrices H t,j, R t,j, A t,j and Q t,j can be different for each target. Extensions to non-linear models is straightforward (EKF, UKF) Target births only occur jointly with an association event: no association to a newborn target, no birth. Equivalently, the state prior is constant until a first measurement is associated, i.e. the target state does not follow the dynamic model before that moment. This very time instant is than identified as target birth. After an association is made, the lifetime of the target is given a probability density p(t d ). A Gamma distribution Γ(α, β) has been chosen for modelling such a probability. Increasing values of α shift the distribution to the right, thus resulting in lower probabilities for small lifetimes, while increasing values of β somehow flatten the distribution. Basically, by modelling deaths, targets which have not been assigned a measure for a certain amount of time are removed. If the last association with target j is made at time τ and at the previous time step t 1 the sampled hypothesis correspond to an alive target, then the probability of having a dead target at current time step t is p(death j ) = P (t d [t 1 τ, t τ] t d t 1 τ). (12) III. SETUP AND MODEL RBMCDA has been tested through a Matlab toolbox [6], which implements all the algorithm s functionalities and eventually provides simulations for testing. Tests have instead been run on real sequences. The dataset that has been considered can be found in [7], together with the ground-truth. It consists in four video sequences corresponding to four different views (i.e. different cameras) of a scene. Six people randomly access Fig. 2. Ground projections of tracked objects ( ) and ground-truth ( ) a 5,5 m x 5.5 m area in a room and walk around quite as much randomly for about 150 seconds at 25 fps. Together with the ground-truth data, homography parameters (coded in 3x3 matrices) for each cameras are supplied, to allow ground plane projections (or, equivalently, to allow a top-view shot). A screenshot of the four video framing is show in Fig. 1, while projections of the tracked objects on the world ground plane are given in Fig 2. Tracked objects are marked with different ( ) symbols (one colour for each camera view), while the ground-truth with a ( ). The model that has been employed is now briefly sketched: the considered state vector is x t = [ x y ẋ ẏ ] T, (13)

4 where x, y, ẋ and ẏ are the horizontal and vertical coordinates and their time variations respectively. The measurement vector is simply given by averaged position measurements z t = [ x y ] T. (14) The transition matrix for the state dynamic evolution is 1 0 t t A t = , (15) where t is simply 1 as only discrete unitary time steps are here considered. The observation matrix is [ ] H t =. (16) The noise R t is taken to be zero-mean Gaussian, with zero offdiagonal correlation terms, while the measurement covariance matrix is of the form Q t = (17) 0 0 σ σ where sigma is an appropriate constant. IV. PERFORMANCE EVALUATION To evaluate the algorithm performances, global traditional metrics, which exploit the information carried by all the particles, could not be employed. The key point is that this leads to an esteem that is corrupted by misassociations between particles and targets, thus being quite meaningless: a particle can in be associated to different targets. Even by considering the particle with higher weight, the association problem persists and its complexity grows exponentially with the number of targets. A novel metric has been therefore designed based on three performance indices which are discussed in the following, namely a fragmentation index I F ; a tracking index I T ; a visibility index I V. a) Fragmentation index: This index focuses on trajectory fragmentation and tries to measure its amount. It consists in the ratio of the difference between the actual number of trajectories T gt and the number of trajectories T generated by the algorithm I f := T T gt T gt. (18) Such an index approaches the optimal value of zero as the number of generated trajectories approaches the number of the real ones; it takes positive values when more trajectories are generated, i.e., roughly speaking, as fragmentation takes place. Negative values of I f basically result from losing a target without retrieving it. b) Tracking index: By means of this index, the number of frames in which target position estimations are effectively related to real data is accounted for (independently of trajectories fragmentation). Equivalently, the index measures the amount of time in which position estimations are supported by ground-truth data, as the fps rate is kept constant. Given the set of all the measured trajectories γ i : ( t) i R 2, let S be the sum of all the supports of γ i, S = i ( t) i and S gt be the same quantity computed for the ground-truth trajectories. The index is defined as I f := S gt S S gt. (19) It approaches zero (optimal value) as fast as the generated trajectories can hook up the when a birth occurs. In fact the former never overstay the latter, as in the case that has been considered real trajectories persist up to the end of the sequence. c) Visibility index: This performance index measures the difference between the actual number of people on the scene (N gt ) and those estimated by the algorithm (N), averaged over all the processed frames N f I v := 1 N f N f (N gt N) i (20) i=1 This index has an optimal value of zero, as the previous. However, its optimization turns out to be extremely hard, as it is strongly affected by frequent and unpredictable occlusions in the scene V. RESULTS Performances have been evaluated by tuning some of the parameters introduced in section II-B, namely P b, which models the birth probability density, and the two parameters α and β, which shape the Gamma function that models the lifetime probability of each target. Two 500-frame scenes with a different intrinsic complexity have been studied.the will be referred to as Scene 1 and 2 (simpler and more complex scene respectively). The three parameters have been varied in turn, while keeping the two other fixed. Table I summarizes some results for Scene 1. The ranges of α and β and P b have been explored systematically to localize local maxima for the tree performance indexes. Low values of α result in short lifetimes, and in fact in trajectories fragmentation, as one can infer from the positive values of I F (Table I(a)). On the other hand, the values of I T and I V do not show monotonicity; instead there seems to be a minima at α = 4. By varying β (Table I(b)), the fragmentation I F remains constant at the optimal null value, while the other two indexes behaviour is quite fuzzy. β = 1 seems to give the best fit. Eventually (Table I(c)), it can be noticed that better performances are obtained with the higher considered value of P b, i.e. P b = I V and I T improve as P b rises: this is basically due to the fact that there is less delay in hooking the target as the birth probability increases. Even higher values of

5 (a) Varying α, β = 1, P b = 0.01 α I F I T I V (b) Varying β, α = 4, P b = 0.01 β I F I T I V (c) Varying P b, α = 4, β = 1 P b I F I T I V TABLE I Scene 1 (a) Varying α, β = 4, P b = 0.1 α I F I T I V (b) Varying β, α = 8, P b = 0.1 β I F I T I V (c) Varying P b, α = 8, β = 5 P b I F I T I V TABLE II Scene 2 P b result in positive values of I F (as for P b = 0.1), as the algorithm generates too many targets. Table II summarizes some results for Scene 1. Here scene is complicated by more targets, resulting in more occlusions and misdetections. The optimal value for α has been found for α = 8 (Table II(a)). It can be noticed that the value of the three indexes show here a kind of local minimum: in particular, the algorithm estimates more trajectories than the real ones for α < 7 (I F > 0) and less for α > 8 (I F < 0). Table II(b) shows how β = 5 represents some kind of asymptote turning point for the values of all the indexes. in particular I V turns positive, meaning that too delayed target deaths result in too many people detected on the scene. From the analysis of Table II(c), it look like that than algorithm cannot avoid fragmenting trajectories in a more complex scene. For Scene 2 a deeper analysis was performed by jointly varying both the indexes α and β, while keeping P b fixed (Figure 3). The vertical coordinates of the plot is z = I F + I T + I V. It can be noticed that z clearly decreases as α and β increase. This is coherent with the above mentioned fact that real trajectories in the scene last until the end of the sequence, i.e. no target exits the scene before the very end. As a matter of fact, high values of α and β allow for long lifetimes. By exploiting the best values of α and β just recovered, the average distance error between real and estimated trajectories has been computed. Its dependence on the number of particles has been also evaluated. Results are given in Table III. It can be noticed that the lowest average error corresponds to n = 10 particles. However, as one can infer from the (negative) value of I F given in the table, this is due to the fact that less trajectories have been generated. Being the number of trajectories equal, better performance is reached as the number of particles increases. This is perfectly coherent with the theory, which states performance improvement on particle Fig. 3. P b = 0.01 increment, especially in noisy conditions. Number of particles Average error (pixels) I F TABLE III AVERAGE ERROR BETWEEN ESTIMATED AND REAL TRAJECTORIES (GROUND-PIXELS) VI. CONCLUSIONS A set-up for RBMCDA testing has been presented. A novel metric based on three indexes has been exploited in order to evaluate the algorithm performance, both qualitatively and quantitatively. Each index accounts for a typical problem in

6 visual tracking. Performance has been evaluated as α, β and P b, which model target birth and death probabilities, vary. Future developments of this work surely include comparative performance analysis with other trackers. Including simple particle filter (non-marginalized filter) evaluation could give evidence to the Rao-Blackwell theorem. REFERENCES [1] A. Doucet, S. Godsill, and C. Andrieu, On Sequential Monte Carlo Sampling Methods for Bayesian Filtering, Statistics and Computing, vol. 10, pp , [Online]. Available: [2] D. B. Reid, An algorithm for tracking multiple targets, IEEE Transactions on Automatic Control, vol. 24, pp , [3] J. Vermaak, S. J. Godsill, and P. Pérez, Monte carlo filtering for multitarget tracking and data association, IEEE Transactions on Aerospace and Electronic Systems, vol. 41, pp , [4] S. Särkkä, A. Vehtari, and J. Lampinen, Rao-blackwellized particle filter for multiple target tracking, Information Fusion Journal, vol. 8, p. 2007, [5] A. Doucet, N. De Freitas, K. Murphy, and S. Russell, Rao-Blackwellised particle filtering for dynamic Bayesian networks, in Proceedings of the Sixteenth Conference on Uncertainty in Artificial Intelligence, 2000, pp [6] LCE Models and Methods: RBMCDA Toolbox for Matlab - TKK - [7] CVLAB, EPFL -

Deterministic Sampling-based Switching Kalman Filtering for Vehicle Tracking

Deterministic Sampling-based Switching Kalman Filtering for Vehicle Tracking Proceedings of the IEEE ITSC 2006 2006 IEEE Intelligent Transportation Systems Conference Toronto, Canada, September 17-20, 2006 WA4.1 Deterministic Sampling-based Switching Kalman Filtering for Vehicle

More information

Introduction to Mobile Robotics Bayes Filter Particle Filter and Monte Carlo Localization

Introduction to Mobile Robotics Bayes Filter Particle Filter and Monte Carlo Localization Introduction to Mobile Robotics Bayes Filter Particle Filter and Monte Carlo Localization Wolfram Burgard, Maren Bennewitz, Diego Tipaldi, Luciano Spinello 1 Motivation Recall: Discrete filter Discretize

More information

Master s thesis tutorial: part III

Master s thesis tutorial: part III for the Autonomous Compliant Research group Tinne De Laet, Wilm Decré, Diederik Verscheure Katholieke Universiteit Leuven, Department of Mechanical Engineering, PMA Division 30 oktober 2006 Outline General

More information

Tracking Groups of Pedestrians in Video Sequences

Tracking Groups of Pedestrians in Video Sequences Tracking Groups of Pedestrians in Video Sequences Jorge S. Marques Pedro M. Jorge Arnaldo J. Abrantes J. M. Lemos IST / ISR ISEL / IST ISEL INESC-ID / IST Lisbon, Portugal Lisbon, Portugal Lisbon, Portugal

More information

Real-time Visual Tracker by Stream Processing

Real-time Visual Tracker by Stream Processing Real-time Visual Tracker by Stream Processing Simultaneous and Fast 3D Tracking of Multiple Faces in Video Sequences by Using a Particle Filter Oscar Mateo Lozano & Kuzahiro Otsuka presented by Piotr Rudol

More information

Tracking in flussi video 3D. Ing. Samuele Salti

Tracking in flussi video 3D. Ing. Samuele Salti Seminari XXIII ciclo Tracking in flussi video 3D Ing. Tutors: Prof. Tullio Salmon Cinotti Prof. Luigi Di Stefano The Tracking problem Detection Object model, Track initiation, Track termination, Tracking

More information

Probability Hypothesis Density filter versus Multiple Hypothesis Tracking

Probability Hypothesis Density filter versus Multiple Hypothesis Tracking Probability Hypothesis Density filter versus Multiple Hypothesis Tracing Kusha Panta a, Ba-Ngu Vo a, Sumeetpal Singh a and Arnaud Doucet b a Co-operative Research Centre for Sensor and Information Processing

More information

A Robust Multiple Object Tracking for Sport Applications 1) Thomas Mauthner, Horst Bischof

A Robust Multiple Object Tracking for Sport Applications 1) Thomas Mauthner, Horst Bischof A Robust Multiple Object Tracking for Sport Applications 1) Thomas Mauthner, Horst Bischof Institute for Computer Graphics and Vision Graz University of Technology, Austria {mauthner,bischof}@icg.tu-graz.ac.at

More information

VEHICLE TRACKING USING ACOUSTIC AND VIDEO SENSORS

VEHICLE TRACKING USING ACOUSTIC AND VIDEO SENSORS VEHICLE TRACKING USING ACOUSTIC AND VIDEO SENSORS Aswin C Sankaranayanan, Qinfen Zheng, Rama Chellappa University of Maryland College Park, MD - 277 {aswch, qinfen, rama}@cfar.umd.edu Volkan Cevher, James

More information

Monte Carlo and Empirical Methods for Stochastic Inference (MASM11/FMS091)

Monte Carlo and Empirical Methods for Stochastic Inference (MASM11/FMS091) Monte Carlo and Empirical Methods for Stochastic Inference (MASM11/FMS091) Magnus Wiktorsson Centre for Mathematical Sciences Lund University, Sweden Lecture 5 Sequential Monte Carlo methods I February

More information

Exploiting A Constellation of Narrowband RF Sensors to Detect and Track Moving Targets

Exploiting A Constellation of Narrowband RF Sensors to Detect and Track Moving Targets Exploiting A Constellation of Narrowband RF Sensors to Detect and Track Moving Targets Chris Kreucher a, J. Webster Stayman b, Ben Shapo a, and Mark Stuff c a Integrity Applications Incorporated 900 Victors

More information

A Tutorial on Particle Filters for Online Nonlinear/Non-Gaussian Bayesian Tracking

A Tutorial on Particle Filters for Online Nonlinear/Non-Gaussian Bayesian Tracking 174 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 50, NO. 2, FEBRUARY 2002 A Tutorial on Particle Filters for Online Nonlinear/Non-Gaussian Bayesian Tracking M. Sanjeev Arulampalam, Simon Maskell, Neil

More information

A Learning Based Method for Super-Resolution of Low Resolution Images

A Learning Based Method for Super-Resolution of Low Resolution Images A Learning Based Method for Super-Resolution of Low Resolution Images Emre Ugur June 1, 2004 emre.ugur@ceng.metu.edu.tr Abstract The main objective of this project is the study of a learning based method

More information

Monte Carlo-based statistical methods (MASM11/FMS091)

Monte Carlo-based statistical methods (MASM11/FMS091) Monte Carlo-based statistical methods (MASM11/FMS091) Jimmy Olsson Centre for Mathematical Sciences Lund University, Sweden Lecture 5 Sequential Monte Carlo methods I February 5, 2013 J. Olsson Monte Carlo-based

More information

Mean-Shift Tracking with Random Sampling

Mean-Shift Tracking with Random Sampling 1 Mean-Shift Tracking with Random Sampling Alex Po Leung, Shaogang Gong Department of Computer Science Queen Mary, University of London, London, E1 4NS Abstract In this work, boosting the efficiency of

More information

Bayesian Statistics: Indian Buffet Process

Bayesian Statistics: Indian Buffet Process Bayesian Statistics: Indian Buffet Process Ilker Yildirim Department of Brain and Cognitive Sciences University of Rochester Rochester, NY 14627 August 2012 Reference: Most of the material in this note

More information

Discrete Frobenius-Perron Tracking

Discrete Frobenius-Perron Tracking Discrete Frobenius-Perron Tracing Barend J. van Wy and Michaël A. van Wy French South-African Technical Institute in Electronics at the Tshwane University of Technology Staatsartillerie Road, Pretoria,

More information

A Comparative Study between SIFT- Particle and SURF-Particle Video Tracking Algorithms

A Comparative Study between SIFT- Particle and SURF-Particle Video Tracking Algorithms A Comparative Study between SIFT- Particle and SURF-Particle Video Tracking Algorithms H. Kandil and A. Atwan Information Technology Department, Faculty of Computer and Information Sciences, Mansoura University,El-Gomhoria

More information

Statistics Graduate Courses

Statistics Graduate Courses Statistics Graduate Courses STAT 7002--Topics in Statistics-Biological/Physical/Mathematics (cr.arr.).organized study of selected topics. Subjects and earnable credit may vary from semester to semester.

More information

Institute of Actuaries of India Subject CT3 Probability and Mathematical Statistics

Institute of Actuaries of India Subject CT3 Probability and Mathematical Statistics Institute of Actuaries of India Subject CT3 Probability and Mathematical Statistics For 2015 Examinations Aim The aim of the Probability and Mathematical Statistics subject is to provide a grounding in

More information

Kristine L. Bell and Harry L. Van Trees. Center of Excellence in C 3 I George Mason University Fairfax, VA 22030-4444, USA kbell@gmu.edu, hlv@gmu.

Kristine L. Bell and Harry L. Van Trees. Center of Excellence in C 3 I George Mason University Fairfax, VA 22030-4444, USA kbell@gmu.edu, hlv@gmu. POSERIOR CRAMÉR-RAO BOUND FOR RACKING ARGE BEARING Kristine L. Bell and Harry L. Van rees Center of Excellence in C 3 I George Mason University Fairfax, VA 22030-4444, USA bell@gmu.edu, hlv@gmu.edu ABSRAC

More information

How To Solve A Sequential Mca Problem

How To Solve A Sequential Mca Problem Monte Carlo-based statistical methods (MASM11/FMS091) Jimmy Olsson Centre for Mathematical Sciences Lund University, Sweden Lecture 6 Sequential Monte Carlo methods II February 3, 2012 Changes in HA1 Problem

More information

Mathieu St-Pierre. Denis Gingras Dr. Ing.

Mathieu St-Pierre. Denis Gingras Dr. Ing. Comparison between the unscented Kalman filter and the extended Kalman filter for the position estimation module of an integrated navigation information system Mathieu St-Pierre Electrical engineering

More information

A Reliability Point and Kalman Filter-based Vehicle Tracking Technique

A Reliability Point and Kalman Filter-based Vehicle Tracking Technique A Reliability Point and Kalman Filter-based Vehicle Tracing Technique Soo Siang Teoh and Thomas Bräunl Abstract This paper introduces a technique for tracing the movement of vehicles in consecutive video

More information

Monte Carlo-based statistical methods (MASM11/FMS091)

Monte Carlo-based statistical methods (MASM11/FMS091) Monte Carlo-based statistical methods (MASM11/FMS091) Magnus Wiktorsson Centre for Mathematical Sciences Lund University, Sweden Lecture 6 Sequential Monte Carlo methods II February 7, 2014 M. Wiktorsson

More information

Accurate and robust image superresolution by neural processing of local image representations

Accurate and robust image superresolution by neural processing of local image representations Accurate and robust image superresolution by neural processing of local image representations Carlos Miravet 1,2 and Francisco B. Rodríguez 1 1 Grupo de Neurocomputación Biológica (GNB), Escuela Politécnica

More information

EE 570: Location and Navigation

EE 570: Location and Navigation EE 570: Location and Navigation On-Line Bayesian Tracking Aly El-Osery 1 Stephen Bruder 2 1 Electrical Engineering Department, New Mexico Tech Socorro, New Mexico, USA 2 Electrical and Computer Engineering

More information

Example: Credit card default, we may be more interested in predicting the probabilty of a default than classifying individuals as default or not.

Example: Credit card default, we may be more interested in predicting the probabilty of a default than classifying individuals as default or not. Statistical Learning: Chapter 4 Classification 4.1 Introduction Supervised learning with a categorical (Qualitative) response Notation: - Feature vector X, - qualitative response Y, taking values in C

More information

Least Squares Estimation

Least Squares Estimation Least Squares Estimation SARA A VAN DE GEER Volume 2, pp 1041 1045 in Encyclopedia of Statistics in Behavioral Science ISBN-13: 978-0-470-86080-9 ISBN-10: 0-470-86080-4 Editors Brian S Everitt & David

More information

On-line Tracking Groups of Pedestrians with Bayesian Networks

On-line Tracking Groups of Pedestrians with Bayesian Networks On-line Tracking Groups of Pedestrians with Bayesian Networks Pedro M. Jorge ISEL / ISR pmj@isel.ipl.pt Jorge S. Marques IST / ISR jsm@isr.ist.utl.pt Arnaldo J. Abrantes ISEL aja@isel.ipl.pt Abstract A

More information

Robust Visual Tracking and Vehicle Classification via Sparse Representation

Robust Visual Tracking and Vehicle Classification via Sparse Representation Robust Visual Tracking and Vehicle Classification via Sparse Representation Xue Mei and Haibin Ling, Member, IEEE Abstract In this paper, we propose a robust visual tracking method by casting tracking

More information

Automatic parameter regulation for a tracking system with an auto-critical function

Automatic parameter regulation for a tracking system with an auto-critical function Automatic parameter regulation for a tracking system with an auto-critical function Daniela Hall INRIA Rhône-Alpes, St. Ismier, France Email: Daniela.Hall@inrialpes.fr Abstract In this article we propose

More information

System Identification for Acoustic Comms.:

System Identification for Acoustic Comms.: System Identification for Acoustic Comms.: New Insights and Approaches for Tracking Sparse and Rapidly Fluctuating Channels Weichang Li and James Preisig Woods Hole Oceanographic Institution The demodulation

More information

STA 4273H: Statistical Machine Learning

STA 4273H: Statistical Machine Learning STA 4273H: Statistical Machine Learning Russ Salakhutdinov Department of Statistics! rsalakhu@utstat.toronto.edu! http://www.cs.toronto.edu/~rsalakhu/ Lecture 6 Three Approaches to Classification Construct

More information

Optimal Stopping in Software Testing

Optimal Stopping in Software Testing Optimal Stopping in Software Testing Nilgun Morali, 1 Refik Soyer 2 1 Department of Statistics, Dokuz Eylal Universitesi, Turkey 2 Department of Management Science, The George Washington University, 2115

More information

Similarity Search and Mining in Uncertain Spatial and Spatio Temporal Databases. Andreas Züfle

Similarity Search and Mining in Uncertain Spatial and Spatio Temporal Databases. Andreas Züfle Similarity Search and Mining in Uncertain Spatial and Spatio Temporal Databases Andreas Züfle Geo Spatial Data Huge flood of geo spatial data Modern technology New user mentality Great research potential

More information

Mapping an Application to a Control Architecture: Specification of the Problem

Mapping an Application to a Control Architecture: Specification of the Problem Mapping an Application to a Control Architecture: Specification of the Problem Mieczyslaw M. Kokar 1, Kevin M. Passino 2, Kenneth Baclawski 1, and Jeffrey E. Smith 3 1 Northeastern University, Boston,

More information

Component Ordering in Independent Component Analysis Based on Data Power

Component Ordering in Independent Component Analysis Based on Data Power Component Ordering in Independent Component Analysis Based on Data Power Anne Hendrikse Raymond Veldhuis University of Twente University of Twente Fac. EEMCS, Signals and Systems Group Fac. EEMCS, Signals

More information

Continuous Time Bayesian Networks for Inferring Users Presence and Activities with Extensions for Modeling and Evaluation

Continuous Time Bayesian Networks for Inferring Users Presence and Activities with Extensions for Modeling and Evaluation Continuous Time Bayesian Networks for Inferring Users Presence and Activities with Extensions for Modeling and Evaluation Uri Nodelman 1 Eric Horvitz Microsoft Research One Microsoft Way Redmond, WA 98052

More information

Real-Time Monitoring of Complex Industrial Processes with Particle Filters

Real-Time Monitoring of Complex Industrial Processes with Particle Filters Real-Time Monitoring of Complex Industrial Processes with Particle Filters Rubén Morales-Menéndez Dept. of Mechatronics and Automation ITESM campus Monterrey Monterrey, NL México rmm@itesm.mx Nando de

More information

ROBUST REAL-TIME ON-BOARD VEHICLE TRACKING SYSTEM USING PARTICLES FILTER. Ecole des Mines de Paris, Paris, France

ROBUST REAL-TIME ON-BOARD VEHICLE TRACKING SYSTEM USING PARTICLES FILTER. Ecole des Mines de Paris, Paris, France ROBUST REAL-TIME ON-BOARD VEHICLE TRACKING SYSTEM USING PARTICLES FILTER Bruno Steux Yotam Abramson Ecole des Mines de Paris, Paris, France Abstract: We describe a system for detection and tracking of

More information

Vision based Vehicle Tracking using a high angle camera

Vision based Vehicle Tracking using a high angle camera Vision based Vehicle Tracking using a high angle camera Raúl Ignacio Ramos García Dule Shu gramos@clemson.edu dshu@clemson.edu Abstract A vehicle tracking and grouping algorithm is presented in this work

More information

Fall detection in the elderly by head tracking

Fall detection in the elderly by head tracking Loughborough University Institutional Repository Fall detection in the elderly by head tracking This item was submitted to Loughborough University's Institutional Repository by the/an author. Citation:

More information

The CUSUM algorithm a small review. Pierre Granjon

The CUSUM algorithm a small review. Pierre Granjon The CUSUM algorithm a small review Pierre Granjon June, 1 Contents 1 The CUSUM algorithm 1.1 Algorithm............................... 1.1.1 The problem......................... 1.1. The different steps......................

More information

Bayesian Image Super-Resolution

Bayesian Image Super-Resolution Bayesian Image Super-Resolution Michael E. Tipping and Christopher M. Bishop Microsoft Research, Cambridge, U.K..................................................................... Published as: Bayesian

More information

Modelling, Extraction and Description of Intrinsic Cues of High Resolution Satellite Images: Independent Component Analysis based approaches

Modelling, Extraction and Description of Intrinsic Cues of High Resolution Satellite Images: Independent Component Analysis based approaches Modelling, Extraction and Description of Intrinsic Cues of High Resolution Satellite Images: Independent Component Analysis based approaches PhD Thesis by Payam Birjandi Director: Prof. Mihai Datcu Problematic

More information

Tutorial on Markov Chain Monte Carlo

Tutorial on Markov Chain Monte Carlo Tutorial on Markov Chain Monte Carlo Kenneth M. Hanson Los Alamos National Laboratory Presented at the 29 th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Technology,

More information

Part 4 fitting with energy loss and multiple scattering non gaussian uncertainties outliers

Part 4 fitting with energy loss and multiple scattering non gaussian uncertainties outliers Part 4 fitting with energy loss and multiple scattering non gaussian uncertainties outliers material intersections to treat material effects in track fit, locate material 'intersections' along particle

More information

A Statistical Framework for Operational Infrasound Monitoring

A Statistical Framework for Operational Infrasound Monitoring A Statistical Framework for Operational Infrasound Monitoring Stephen J. Arrowsmith Rod W. Whitaker LA-UR 11-03040 The views expressed here do not necessarily reflect the views of the United States Government,

More information

Inference of Probability Distributions for Trust and Security applications

Inference of Probability Distributions for Trust and Security applications Inference of Probability Distributions for Trust and Security applications Vladimiro Sassone Based on joint work with Mogens Nielsen & Catuscia Palamidessi Outline 2 Outline Motivations 2 Outline Motivations

More information

Vehicle Tracking in Occlusion and Clutter

Vehicle Tracking in Occlusion and Clutter Vehicle Tracking in Occlusion and Clutter by KURTIS NORMAN MCBRIDE A thesis presented to the University of Waterloo in fulfilment of the thesis requirement for the degree of Master of Applied Science in

More information

A SURVEY ON CONTINUOUS ELLIPTICAL VECTOR DISTRIBUTIONS

A SURVEY ON CONTINUOUS ELLIPTICAL VECTOR DISTRIBUTIONS A SURVEY ON CONTINUOUS ELLIPTICAL VECTOR DISTRIBUTIONS Eusebio GÓMEZ, Miguel A. GÓMEZ-VILLEGAS and J. Miguel MARÍN Abstract In this paper it is taken up a revision and characterization of the class of

More information

Lecture 3: Linear methods for classification

Lecture 3: Linear methods for classification Lecture 3: Linear methods for classification Rafael A. Irizarry and Hector Corrada Bravo February, 2010 Today we describe four specific algorithms useful for classification problems: linear regression,

More information

JPEG compression of monochrome 2D-barcode images using DCT coefficient distributions

JPEG compression of monochrome 2D-barcode images using DCT coefficient distributions Edith Cowan University Research Online ECU Publications Pre. JPEG compression of monochrome D-barcode images using DCT coefficient distributions Keng Teong Tan Hong Kong Baptist University Douglas Chai

More information

Tracking performance evaluation on PETS 2015 Challenge datasets

Tracking performance evaluation on PETS 2015 Challenge datasets Tracking performance evaluation on PETS 2015 Challenge datasets Tahir Nawaz, Jonathan Boyle, Longzhen Li and James Ferryman Computational Vision Group, School of Systems Engineering University of Reading,

More information

EM Clustering Approach for Multi-Dimensional Analysis of Big Data Set

EM Clustering Approach for Multi-Dimensional Analysis of Big Data Set EM Clustering Approach for Multi-Dimensional Analysis of Big Data Set Amhmed A. Bhih School of Electrical and Electronic Engineering Princy Johnson School of Electrical and Electronic Engineering Martin

More information

Introduction to General and Generalized Linear Models

Introduction to General and Generalized Linear Models Introduction to General and Generalized Linear Models General Linear Models - part I Henrik Madsen Poul Thyregod Informatics and Mathematical Modelling Technical University of Denmark DK-2800 Kgs. Lyngby

More information

Lecture 9: Introduction to Pattern Analysis

Lecture 9: Introduction to Pattern Analysis Lecture 9: Introduction to Pattern Analysis g Features, patterns and classifiers g Components of a PR system g An example g Probability definitions g Bayes Theorem g Gaussian densities Features, patterns

More information

Understanding Purposeful Human Motion

Understanding Purposeful Human Motion M.I.T Media Laboratory Perceptual Computing Section Technical Report No. 85 Appears in Fourth IEEE International Conference on Automatic Face and Gesture Recognition Understanding Purposeful Human Motion

More information

1 Prior Probability and Posterior Probability

1 Prior Probability and Posterior Probability Math 541: Statistical Theory II Bayesian Approach to Parameter Estimation Lecturer: Songfeng Zheng 1 Prior Probability and Posterior Probability Consider now a problem of statistical inference in which

More information

Simultaneous Gamma Correction and Registration in the Frequency Domain

Simultaneous Gamma Correction and Registration in the Frequency Domain Simultaneous Gamma Correction and Registration in the Frequency Domain Alexander Wong a28wong@uwaterloo.ca William Bishop wdbishop@uwaterloo.ca Department of Electrical and Computer Engineering University

More information

Simple Linear Regression Inference

Simple Linear Regression Inference Simple Linear Regression Inference 1 Inference requirements The Normality assumption of the stochastic term e is needed for inference even if it is not a OLS requirement. Therefore we have: Interpretation

More information

STRATEGIC CAPACITY PLANNING USING STOCK CONTROL MODEL

STRATEGIC CAPACITY PLANNING USING STOCK CONTROL MODEL Session 6. Applications of Mathematical Methods to Logistics and Business Proceedings of the 9th International Conference Reliability and Statistics in Transportation and Communication (RelStat 09), 21

More information

Bayesian networks - Time-series models - Apache Spark & Scala

Bayesian networks - Time-series models - Apache Spark & Scala Bayesian networks - Time-series models - Apache Spark & Scala Dr John Sandiford, CTO Bayes Server Data Science London Meetup - November 2014 1 Contents Introduction Bayesian networks Latent variables Anomaly

More information

Analysis of Bayesian Dynamic Linear Models

Analysis of Bayesian Dynamic Linear Models Analysis of Bayesian Dynamic Linear Models Emily M. Casleton December 17, 2010 1 Introduction The main purpose of this project is to explore the Bayesian analysis of Dynamic Linear Models (DLMs). The main

More information

Model-based Synthesis. Tony O Hagan

Model-based Synthesis. Tony O Hagan Model-based Synthesis Tony O Hagan Stochastic models Synthesising evidence through a statistical model 2 Evidence Synthesis (Session 3), Helsinki, 28/10/11 Graphical modelling The kinds of models that

More information

Statistical Machine Learning from Data

Statistical Machine Learning from Data Samy Bengio Statistical Machine Learning from Data 1 Statistical Machine Learning from Data Gaussian Mixture Models Samy Bengio IDIAP Research Institute, Martigny, Switzerland, and Ecole Polytechnique

More information

Robust Infrared Vehicle Tracking across Target Pose Change using L 1 Regularization

Robust Infrared Vehicle Tracking across Target Pose Change using L 1 Regularization Robust Infrared Vehicle Tracking across Target Pose Change using L 1 Regularization Haibin Ling 1, Li Bai, Erik Blasch 3, and Xue Mei 4 1 Computer and Information Science Department, Temple University,

More information

Tracking Pedestrians using Local Spatio-temporal Motion Patterns in Extremely Crowded Scenes

Tracking Pedestrians using Local Spatio-temporal Motion Patterns in Extremely Crowded Scenes 1 Submitted to IEEE Trans. on Pattern Analysis and Machine Intelligence Regular Paper Tracking Pedestrians using Local Spatio-temporal Motion Patterns in Extremely Crowded Scenes Louis Kratz and Ko Nishino

More information

Spatial Statistics Chapter 3 Basics of areal data and areal data modeling

Spatial Statistics Chapter 3 Basics of areal data and areal data modeling Spatial Statistics Chapter 3 Basics of areal data and areal data modeling Recall areal data also known as lattice data are data Y (s), s D where D is a discrete index set. This usually corresponds to data

More information

Current Standard: Mathematical Concepts and Applications Shape, Space, and Measurement- Primary

Current Standard: Mathematical Concepts and Applications Shape, Space, and Measurement- Primary Shape, Space, and Measurement- Primary A student shall apply concepts of shape, space, and measurement to solve problems involving two- and three-dimensional shapes by demonstrating an understanding of:

More information

A Movement Tracking Management Model with Kalman Filtering Global Optimization Techniques and Mahalanobis Distance

A Movement Tracking Management Model with Kalman Filtering Global Optimization Techniques and Mahalanobis Distance Loutraki, 21 26 October 2005 A Movement Tracking Management Model with ing Global Optimization Techniques and Raquel Ramos Pinho, João Manuel R. S. Tavares, Miguel Velhote Correia Laboratório de Óptica

More information

Wii Remote Calibration Using the Sensor Bar

Wii Remote Calibration Using the Sensor Bar Wii Remote Calibration Using the Sensor Bar Alparslan Yildiz Abdullah Akay Yusuf Sinan Akgul GIT Vision Lab - http://vision.gyte.edu.tr Gebze Institute of Technology Kocaeli, Turkey {yildiz, akay, akgul}@bilmuh.gyte.edu.tr

More information

Kalman Filter Applied to a Active Queue Management Problem

Kalman Filter Applied to a Active Queue Management Problem IOSR Journal of Electrical and Electronics Engineering (IOSR-JEEE) e-issn: 2278-1676,p-ISSN: 2320-3331, Volume 9, Issue 4 Ver. III (Jul Aug. 2014), PP 23-27 Jyoti Pandey 1 and Prof. Aashih Hiradhar 2 Department

More information

The Scientific Data Mining Process

The Scientific Data Mining Process Chapter 4 The Scientific Data Mining Process When I use a word, Humpty Dumpty said, in rather a scornful tone, it means just what I choose it to mean neither more nor less. Lewis Carroll [87, p. 214] In

More information

Tracking and Recognition in Sports Videos

Tracking and Recognition in Sports Videos Tracking and Recognition in Sports Videos Mustafa Teke a, Masoud Sattari b a Graduate School of Informatics, Middle East Technical University, Ankara, Turkey mustafa.teke@gmail.com b Department of Computer

More information

Distributed computing of failure probabilities for structures in civil engineering

Distributed computing of failure probabilities for structures in civil engineering Distributed computing of failure probabilities for structures in civil engineering Andrés Wellmann Jelic, University of Bochum (andres.wellmann@rub.de) Matthias Baitsch, University of Bochum (matthias.baitsch@rub.de)

More information

Practical Tour of Visual tracking. David Fleet and Allan Jepson January, 2006

Practical Tour of Visual tracking. David Fleet and Allan Jepson January, 2006 Practical Tour of Visual tracking David Fleet and Allan Jepson January, 2006 Designing a Visual Tracker: What is the state? pose and motion (position, velocity, acceleration, ) shape (size, deformation,

More information

CCNY. BME I5100: Biomedical Signal Processing. Linear Discrimination. Lucas C. Parra Biomedical Engineering Department City College of New York

CCNY. BME I5100: Biomedical Signal Processing. Linear Discrimination. Lucas C. Parra Biomedical Engineering Department City College of New York BME I5100: Biomedical Signal Processing Linear Discrimination Lucas C. Parra Biomedical Engineering Department CCNY 1 Schedule Week 1: Introduction Linear, stationary, normal - the stuff biology is not

More information

Machine Learning and Pattern Recognition Logistic Regression

Machine Learning and Pattern Recognition Logistic Regression Machine Learning and Pattern Recognition Logistic Regression Course Lecturer:Amos J Storkey Institute for Adaptive and Neural Computation School of Informatics University of Edinburgh Crichton Street,

More information

Supplement to Call Centers with Delay Information: Models and Insights

Supplement to Call Centers with Delay Information: Models and Insights Supplement to Call Centers with Delay Information: Models and Insights Oualid Jouini 1 Zeynep Akşin 2 Yves Dallery 1 1 Laboratoire Genie Industriel, Ecole Centrale Paris, Grande Voie des Vignes, 92290

More information

Traffic Flow Monitoring in Crowded Cities

Traffic Flow Monitoring in Crowded Cities Traffic Flow Monitoring in Crowded Cities John A. Quinn and Rose Nakibuule Faculty of Computing & I.T. Makerere University P.O. Box 7062, Kampala, Uganda {jquinn,rnakibuule}@cit.mak.ac.ug Abstract Traffic

More information

Chapter 14 Managing Operational Risks with Bayesian Networks

Chapter 14 Managing Operational Risks with Bayesian Networks Chapter 14 Managing Operational Risks with Bayesian Networks Carol Alexander This chapter introduces Bayesian belief and decision networks as quantitative management tools for operational risks. Bayesian

More information

An Introduction to the Kalman Filter

An Introduction to the Kalman Filter An Introduction to the Kalman Filter Greg Welch 1 and Gary Bishop 2 TR 95041 Department of Computer Science University of North Carolina at Chapel Hill Chapel Hill, NC 275993175 Updated: Monday, July 24,

More information

Bayesian tracking of elongated structures in 3D images

Bayesian tracking of elongated structures in 3D images Bayesian tracking of elongated structures in 3D images Michiel Schaap, Ihor Smal, Coert Metz, Theo van Walsum, and Wiro Niessen Biomedical Imaging Group Rotterdam, Departments of Radiology and Medical

More information

Automatic Traffic Estimation Using Image Processing

Automatic Traffic Estimation Using Image Processing Automatic Traffic Estimation Using Image Processing Pejman Niksaz Science &Research Branch, Azad University of Yazd, Iran Pezhman_1366@yahoo.com Abstract As we know the population of city and number of

More information

Message-passing sequential detection of multiple change points in networks

Message-passing sequential detection of multiple change points in networks Message-passing sequential detection of multiple change points in networks Long Nguyen, Arash Amini Ram Rajagopal University of Michigan Stanford University ISIT, Boston, July 2012 Nguyen/Amini/Rajagopal

More information

Data Modeling & Analysis Techniques. Probability & Statistics. Manfred Huber 2011 1

Data Modeling & Analysis Techniques. Probability & Statistics. Manfred Huber 2011 1 Data Modeling & Analysis Techniques Probability & Statistics Manfred Huber 2011 1 Probability and Statistics Probability and statistics are often used interchangeably but are different, related fields

More information

An Autonomous Agent for Supply Chain Management

An Autonomous Agent for Supply Chain Management In Gedas Adomavicius and Alok Gupta, editors, Handbooks in Information Systems Series: Business Computing, Emerald Group, 2009. An Autonomous Agent for Supply Chain Management David Pardoe, Peter Stone

More information

11. Time series and dynamic linear models

11. Time series and dynamic linear models 11. Time series and dynamic linear models Objective To introduce the Bayesian approach to the modeling and forecasting of time series. Recommended reading West, M. and Harrison, J. (1997). models, (2 nd

More information

Visual Tracking. Frédéric Jurie LASMEA. CNRS / Université Blaise Pascal. France. jurie@lasmea.univ-bpclermoaddressnt.fr

Visual Tracking. Frédéric Jurie LASMEA. CNRS / Université Blaise Pascal. France. jurie@lasmea.univ-bpclermoaddressnt.fr Visual Tracking Frédéric Jurie LASMEA CNRS / Université Blaise Pascal France jurie@lasmea.univ-bpclermoaddressnt.fr What is it? Visual tracking is the problem of following moving targets through an image

More information

NEW YORK STATE TEACHER CERTIFICATION EXAMINATIONS

NEW YORK STATE TEACHER CERTIFICATION EXAMINATIONS NEW YORK STATE TEACHER CERTIFICATION EXAMINATIONS TEST DESIGN AND FRAMEWORK September 2014 Authorized for Distribution by the New York State Education Department This test design and framework document

More information

A PHD filter for tracking multiple extended targets using random matrices

A PHD filter for tracking multiple extended targets using random matrices A PHD filter for tracing multiple extended targets using random matrices Karl Granström and Umut Orguner Linöping University Post Print N.B.: When citing this wor, cite the original article. IEEE. Personal

More information

Linear Classification. Volker Tresp Summer 2015

Linear Classification. Volker Tresp Summer 2015 Linear Classification Volker Tresp Summer 2015 1 Classification Classification is the central task of pattern recognition Sensors supply information about an object: to which class do the object belong

More information

Crater detection with segmentation-based image processing algorithm

Crater detection with segmentation-based image processing algorithm Template reference : 100181708K-EN Crater detection with segmentation-based image processing algorithm M. Spigai, S. Clerc (Thales Alenia Space-France) V. Simard-Bilodeau (U. Sherbrooke and NGC Aerospace,

More information

Using Symmetric State Transformations for Multi-Target Tracking

Using Symmetric State Transformations for Multi-Target Tracking Using Symmetric State Transformations for Multi-Target Tracking Marcus Baum and Uwe D. Hanebeck Intelligent Sensor-Actuator-Systems Laboratory ISAS Institute for Anthropomatics Karlsruhe Institute of Technology

More information

Static Environment Recognition Using Omni-camera from a Moving Vehicle

Static Environment Recognition Using Omni-camera from a Moving Vehicle Static Environment Recognition Using Omni-camera from a Moving Vehicle Teruko Yata, Chuck Thorpe Frank Dellaert The Robotics Institute Carnegie Mellon University Pittsburgh, PA 15213 USA College of Computing

More information

Enhancing the SNR of the Fiber Optic Rotation Sensor using the LMS Algorithm

Enhancing the SNR of the Fiber Optic Rotation Sensor using the LMS Algorithm 1 Enhancing the SNR of the Fiber Optic Rotation Sensor using the LMS Algorithm Hani Mehrpouyan, Student Member, IEEE, Department of Electrical and Computer Engineering Queen s University, Kingston, Ontario,

More information

Prediction of Heart Disease Using Naïve Bayes Algorithm

Prediction of Heart Disease Using Naïve Bayes Algorithm Prediction of Heart Disease Using Naïve Bayes Algorithm R.Karthiyayini 1, S.Chithaara 2 Assistant Professor, Department of computer Applications, Anna University, BIT campus, Tiruchirapalli, Tamilnadu,

More information

Rao-Blackwellized Marginal Particle Filtering for Multiple Object Tracking in Molecular Bioimaging

Rao-Blackwellized Marginal Particle Filtering for Multiple Object Tracking in Molecular Bioimaging Rao-Blackwellized Marginal Particle Filtering for Multiple Object Tracking in Molecular Bioimaging Ihor Smal 1, Katharina Draegestein 2, Niels Galjart 2, Wiro Niessen 1, Erik Meijering 1 1 Departments

More information